Hello! We are a team Dev’s Battle. In this post, we will talk about how we integrated Chat GPT (and continue to integrate) into our product (an MMO RPG telegram game for beginner developers).
For the past two weeks, the Internet has been full of stories about the almighty Chat GPT and its capabilities. So we decided not to stand aside and tell how we found application for such a promising technology, and saved hundreds of man hours for the team.
How it all began
Our product is a telegram game where developers can solve puzzles in their programming languages, earn points for it and battle with other characters.
Recently (two months) the whole team has been actively involved in the process of expanding the functionality, fixing bugs and much more. During this time, our database of several thousand questions and tasks, carefully selected by our developers, began to bother the players, the questions began to be repeated and the players began to complain about it.
This moment coincided with our first attempts to reach an English-speaking audience, where we came to a similar problem. After translating 4000 tasks by auto-translation into English, we realized with horror that such a quality of tasks would definitely not work. For example, auto-translation began to directly translate the names of commands or methods in Python and Java, which made these tasks not only meaningless, but also simply wrong.
We had to look for a solution, and as often happens, the solution found us by itself. And his name is Chat GPT))
Discovery and insight
We will not lie and say that this idea came to us by itself. The solution was proposed by a member of our small game community, who apparently was already tired of solving the same type of tasks.
We immediately realized that it was worth a try and launched a series of our experiments with a neuron. We immediately understood one important thing, the more precisely you formulate the task for her, the better your result will be.
Therefore, almost immediately from the tasks “Come up with 100 Python problems” we switched to the format:
“Come up with 100 unique and challenging Python problems in a 4-choice quiz format, underline the correct answer and provide an extended explanation”
We also actively began to use Chat GPT to generate tasks in English, since our team does not have a native speaker, and translations take a significant amount of time.
To date, Chat GPT has already generated over 300 tasks for us. The only disadvantages that we found 1) after n-requests, such puzzles begin to repeat themselves and 2) puzzles need to be interrupted into our database by hand (we found something to complain about, the reader will say)
Chat GPT and the future for our project
Now we are actively working on options for working with the Chat GPT API so that we can directly connect it to our database and set up automatic completion of questions.
We also hope that such a solution will allow us to increase the performance of the neuron. Now the service often hangs (as they say on the Internet – due to the influx of bloggers) and does not allow us to generate large samples of questions.
The price of the solution, as it seems to us, is more than justified. And for such cool things it’s definitely worth paying for. In any case, it will be cheaper than hiring a bunch of people for the project.
We believe that neurons are the future, which is increasingly flowing into our present. We are sure that the application for Chat GPT can be found in a bunch of business processes, for example, writing articles or SMM. Perhaps we will also ask you to write the next article for us Chat GPT))
In this article, we did not try to evaluate Chat GPT, make a business analysis of the technology and show its application in full glory. Just wanted to share our case for the community. Therefore, do not judge strictly.
PSIf you want to try our game yourself – here link. We welcome your comments, ideas and suggestions.